Add Foundry jobs and MLflow surfaces#42790
Open
nlydick wants to merge 35 commits intoAzure:feature/foundry-stagingfrom
Open
Add Foundry jobs and MLflow surfaces#42790nlydick wants to merge 35 commits intoAzure:feature/foundry-stagingfrom
nlydick wants to merge 35 commits intoAzure:feature/foundry-stagingfrom
Conversation
Next Steps to MergeImportant checks have failed. As of today they are not blocking this PR, but in near future they may.Addressing the following failures is highly recommended:
Comment generated by summarize-checks workflow run. |
API Change CheckAPIView identified API level changes in this PR and created the following API reviews
Comment generated by After APIView workflow run. |
4279cbb to
49f2e2e
Compare
Add the Foundry jobs TypeSpec surface, generated OpenAPI snapshots, and examples. Model training creation now submits with POST /jobs and Operation-Id idempotency, while SDK customization keeps the convenience operation under beta models. Expose job attempts, artifacts, metrics, and outputs under /jobs, including attempt-pinned views. Add MLflow tracking and registry compatibility routes under /mlflow/api/2.0/mlflow. Authored-by: GitHub Copilot for VS Code 0.45.1 Model: GitHub Copilot Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
49f2e2e to
245f2c4
Compare
Update the Foundry training job TypeSpec, generated OpenAPI output, and examples from the latest PR bundle. Rename the generated examples from Jobs_* to TrainingJobs_* to match the updated operation surface. Authored-by: GitHub Copilot for VS Code v0.45.1 Model: GitHub Copilot (github-copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Add a typed computeDetails schema for training job run attempts so SDKs expose resolved placement fields such as VM size, instance count, GPU count, region, and backend-specific properties. Authored-by: GitHub Copilot for VS Code 0.45.1 Model: GitHub Copilot (github-copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Commit the generated virtual-public-preview OpenAPI output as currently produced in the PR branch. Authored-by: GitHub Copilot for VS Code 0.45.1 Model: GitHub Copilot (github-copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Add the Foundry training job recipe payload to the TypeSpec and expose the data-plane operation polling routes used by MFE/APIM. Refresh the generated OpenAPI and create examples so SDKs see the declarative training shape. Authored-by: GitHub Copilot for VS Code 0.45.1 Model: GitHub Copilot (github-copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Add Operation-Location headers to training job delete, cancel, and accepted operation polling responses so generated SDK pollers can follow the status route while Location remains the result route. Authored-by: GitHub Copilot for VS Code v0.45.1 Model: GitHub Copilot (copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Model training job service endpoints as a jobs preview operation with an optional nodeId query parameter. Regenerate the v1 and virtual-public-preview OpenAPI outputs for the new services response shape. Authored-by: GitHub Copilot for VS Code v0.45.1 Model: GitHub Copilot (copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Add SDK-visible trainingJobs prefix content-info operations and regenerate Foundry OpenAPI output for default artifact downloads. Authored-by: GitHub Copilot for VS Code v0.45.1 Model: GitHub Copilot Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Use a create-specific job body model and remove unsupported child-list paging parameters from the new trainingJobs surface. Authored-by: GitHub Copilot for VS Code v0.45.1 Model: GitHub Copilot Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Keep generated OpenAPI JSON out of this PR's swagger validation path, add local Foundry readmes for Avocado, and suppress the existing Foundry TypeSpec validation integration gap. Authored-by: GitHub Copilot for VS Code v0.45.1 Model: GitHub Copilot Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Add the AutoRest markdown marker required by Avocado to the Foundry readmes without adding new generated swagger inputs. Authored-by: GitHub Copilot for VS Code v0.45.1 Model: GitHub Copilot Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Complete the project MLflow compatibility surface by adding missing route variants and matching request shapes used by the Vienna MLflow 2.0 controllers. Authored-by: GitHub Copilot for VS Code v0.45.1 Model: GitHub Copilot Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Update the Foundry training job REST resource and generated artifacts to use training_jobs as the wire path and resource segment. Authored-by: GitHub Copilot for VS Code 0.45.1 Model: GitHub Copilot (copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Add model-scoped managed training APIs, operation polling, stricter training configuration contracts, and cluster-scoped interactive Ray job creation. Regenerate Foundry OpenAPI output and examples for the new routes. Authored-by: GitHub Copilot for VS Code 0.45.1 Model: GitHub Copilot (copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Add Ray client remote-access fields to the interactive Ray request and Ray distribution contracts. Regenerate Foundry OpenAPI output and update the interactive Ray example. Authored-by: GitHub Copilot for VS Code 0.45.1 Model: GitHub Copilot (copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Improve SDK/OpenAPI documentation for algorithm-specific dataset column mappings and required runtime fields. Authored-by: GitHub Copilot for VS Code 0.45.1 Model: GitHub Copilot (copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Rename the interactive Ray cluster action to start_ray_session, add stop_ray_session, refresh OpenAPI/examples, and expose missing low-level training column mappings for response, label, and score. Authored-by: GitHub Copilot for VS Code 0.45.1 Model: GitHub Copilot (copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Rename the attempt-pinned training job child routes from runs to attempts so the public path and schema match the backing run attempt semantics. Authored-by: GitHub Copilot for VS Code 0.45.1 Model: GitHub Copilot (copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Move training job artifact and metric routes for the current attempt under attempts/latest so the public surface is explicit about attempt semantics. Authored-by: GitHub Copilot for VS Code 0.45.1 Model: GitHub Copilot (copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Align training job attempt routes with the implementation contract. Use integer attempt IDs so latest remains an explicit reserved route segment. Authored-by: GitHub Copilot for VS Code 0.45.1 Model: GitHub Copilot (copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Move beta relocation customizations to the renamed training-job attempt operation groups and add C# client projection names for clusters, models, and training jobs. Authored-by: GitHub Copilot for VS Code v0.45.1 Model: GitHub Copilot (copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Collapse managed model training examples and OpenAPI to the typed create-training-job shape. Remove the cluster Ray action and document Ray through generic command training jobs. Authored-by: GitHub Copilot for VS Code 0.45.1 Model: GitHub Copilot (copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
trrwilson
reviewed
May 6, 2026
glecaros
reviewed
May 7, 2026
glecaros
reviewed
May 7, 2026
glecaros
reviewed
May 7, 2026
glecaros
reviewed
May 7, 2026
glecaros
reviewed
May 7, 2026
glecaros
reviewed
May 7, 2026
glecaros
reviewed
May 7, 2026
glecaros
reviewed
May 7, 2026
glecaros
reviewed
May 7, 2026
glecaros
reviewed
May 7, 2026
glecaros
reviewed
May 7, 2026
glecaros
reviewed
May 7, 2026
glecaros
reviewed
May 7, 2026
glecaros
reviewed
May 7, 2026
glecaros
reviewed
May 7, 2026
glecaros
reviewed
May 7, 2026
glecaros
reviewed
May 7, 2026
Address review feedback on the Foundry training jobs data-plane surface. Rename the jobs TypeSpec folder and preview flag to training_jobs, align wire names with snake_case, use common pagination patterns, collapse latest attempt routes, and regenerate OpenAPI/examples. Authored-by: GitHub Copilot for VS Code Model: GitHub Copilot Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Add an SDK-only InteractiveRayJob input shape for creating Ray-backed command jobs through the existing training jobs create operation. Limit the convenience framework enum to Slime and VERL, and remove stale SDK relocation hooks for artifact and metric operation groups that no longer exist. Authored-by: GitHub Copilot for VS Code v0.45.1 Model: GitHub Copilot (copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Move the preview training jobs SDK surface from beta.model.training_jobs to the direct beta.training_jobs client, matching the surrounding preview resource clients. Relocate showServices with the rest of the training jobs operations so the generated SDK no longer splits the resource surface. Authored-by: GitHub Copilot for VS Code v0.45.1 Model: GitHub Copilot (copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Relocate the remaining required-preview training job operation routes into the beta.training_jobs SDK client so artifact prefix helpers and polling routes do not generate stray clients. Authored-by: GitHub Copilot for VS Code v0.45.1 Model: GitHub Copilot (copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Relocate MLflow operation groups into the beta SDK client surface while preserving the existing MLflow wire routes. Keep the MLflow routes grouped with beta SDK APIs without adding a preview feature header requirement or changing MFE/APIM route contracts. Authored-by: GitHub Copilot for VS Code v0.45.1 Model: GitHub Copilot (copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Flatten training job create and read schemas by spreading command job fields onto the top-level resource shape. Regenerate OpenAPI and examples for the flattened data-plane payloads. Authored-by: GitHub Copilot for VS Code v0.45.1 Model: GitHub Copilot (github-copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Relocate managed model training create overloads and polling results to the beta SDK client surface. Keep the client import grouping aligned with the preview feature requirement for these routes. Authored-by: GitHub Copilot for VS Code v0.45.1 Model: GitHub Copilot (github-copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Expose a single beta model-training LRO method in Python and JavaScript while keeping algorithm-specific request bodies as job types. Hide algorithm-specific create overload operations from Python and JavaScript SDK generation. Authored-by: GitHub Copilot for VS Code v0.45.1 Model: GitHub Copilot (github-copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Expose next_after and UX-visible attempt, artifact, and metric metadata on training_jobs while preserving existing route shapes. Regenerate Foundry OpenAPI. Authored-by: GitHub Copilot for VS Code 0.45.1 Model: GitHub Copilot (github-copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Expose include=latest_attempt on training job get, document the read-only latest_attempt projection, and keep examples plus generated OpenAPI in sync. Authored-by: GitHub Copilot for VS Code 0.45.1 Model: GitHub Copilot (unknown) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Use explicit required-preview operations for top-level training job member routes so generated OpenAPI uses job_name like the child routes. Regenerate OpenAPI for v1 and virtual-public-preview. Authored-by: GitHub Copilot for VS Code Model: GitHub Copilot (copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Removes unsupported before cursors from training job list surfaces while keeping order where MFE can implement it. Authored-by: GitHub Copilot for VS Code Model: GitHub Copilot (copilot) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Document expansion-style include values for GET training_jobs/{job_name} and add read-only response fields for RunHistory-backed job details.
Regenerate Foundry OpenAPI outputs from TypeSpec.
Authored-by: GitHub Copilot for VS Code
Model: GitHub Copilot (copilot)
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
/jobsroute root.POST /jobsplusOperation-Ididempotency, while relocating the SDK convenience operation to the betamodelsclient ascreate_training_job/createTrainingJob.jobsand expose job child surfaces for attempts, artifacts, metrics, outputs, and attempt-pinned artifact/metric/output views./mlflow/api/2.0/mlflowfor experiments, runs, metrics, artifacts, registered models, and model versions.Validation
git diff --check -- specification/ai-foundry/data-plane/Foundrynpx tsp compile specification/ai-foundry/data-plane/Foundry --emit @typespec/openapi3npx tsp compile specification/ai-foundry/data-plane/Foundry/client.tsp --no-emitsrc/agents/routes.tspfor unnamed unions.